Adaptive receivers, such as a normalized least mean square (NLMS) equalizer used in wireless transmit/receive units (WTRUs) and base stations, optimize their associated filter tap values through an iterative procedure that requires multiple iterations to near convergence. The tap values converge as time passes to a minimum mean square error (MMSE) solution used to perform channel estimation.
An NLMS receiver includes an equalizer having an equalizer filter which is continually in the process of converging as it tries to track a time-varying channel. The more complex it is to track the channel, the further the tap values of the equalizer will be from convergence. Generally, faster channels (i.e., channel states that evolve rapidly) are difficult for the equalizer to track. Residual automatic frequency control (AFC) errors in the baseband input into the equalizer cause channels to appear faster than they really are. The increase in the apparent speed of the channel can only be partially mitigated by increasing the step-size of an NLMS algorithm implemented by the NLMS receiver. The increased step-size allows the equalizer filter to more accurately track “fast” channels, but it also increases errors in the MMSE solution which cause degradation in the performance of the receiver.
Receivers that employ channel estimation are also degraded by residual AFC errors. Since the bandwidth of the appropriate equalizer filter used in channel estimation is a function of the apparent speed of the channel, large AFC errors force the use of wide-band filters that do not efficiently suppress noise, thus leading to less accurate channel estimates. A simple solution is desired to suppress the residual AFC errors.